When ESG Disclosure Pays but Markets Look Away: Evidence from Vietnamese Commercial Banks

1. Introduction

Environmental, social, and governance (ESG) disclosure has moved from a peripheral reporting practice to a central component of corporate accountability. For banks, the stakes are distinctive: as the primary intermediaries channeling capital across the economy, their lending and investment decisions carry system-wide environmental and social consequences, and their transparency shapes the confidence of depositors, investors, regulators, and the public (Wu & Shen, 2013; Shen et al., 2016; Nizam et al., 2019; Azmi et al., 2021). A growing body of research has therefore asked whether banks that disclose more ESG information also perform better financially. Yet the evidence remains strikingly inconsistent: studies report positive (Buallay et al., 2021; Gangwani, 2024), negative (Forgione et al., 2020), and insignificant or non-linear relationships (Bătae et al., 2020), and the meta-analytic literature concludes that the ESG–performance nexus is highly context-dependent (Friede et al., 2015; Whelan et al., 2021).

This inconsistency is amplified in emerging markets, where the value of transparency is arguably greatest—because regulation, corporate governance, and information environments are weakest—but where empirical evidence is thinnest (Shakil et al., 2019; Azmi et al., 2021). Emerging-market banks operate in settings of pronounced information asymmetry, in which credible non-financial disclosure can materially reduce uncertainty for stakeholders. At the same time, sustainability reporting in these markets is largely voluntary, uneven, and only loosely tied to strategy, raising the question of whether disclosure translates into measurable financial outcomes at all.

Vietnam offers a compelling and under-studied laboratory. Its banking sector is bank-dominated, rapidly growing, and increasingly steered toward sustainability by policy: the State Bank of Vietnam’s green-banking and green-credit directives, the Ministry of Finance’s sustainability-reporting guidance, and the country’s net-zero-by-2050 commitment at COP26 have jointly elevated ESG onto the sector’s agenda. Prior Vietnamese evidence is valuable but limited to short, recent windows and to accounting measures alone (e.g., Loan et al., 2024, over 2018–2022), leaving open how disclosure has evolved over a longer horizon and whether it is reflected in market valuation.

This study addresses these gaps. Using a 16-year unbalanced panel of Vietnamese commercial banks (2010–2025), we build a voluntary ESG disclosure index through content analysis across the three pillars and examine its association with both accounting performance (ROA, ROE) and market performance (Tobin’s Q, price-to-book). We ask two questions: (i) Does voluntary ESG disclosure improve bank financial performance? and (ii) Does the effect differ between the accounting and market dimensions of performance?

Our results reveal a clear asymmetry. Voluntary ESG disclosure is positively and significantly related to accounting performance—most strongly to ROE—and its effect on ROA becomes significant once the collinearity between disclosure, bank size, and age is addressed. In sharp contrast, ESG disclosure bears no significant relationship to either market measure. In other words, transparency is rewarded in the books but not (yet) in the market: the Vietnamese capital market has not fully priced sustainability disclosure into bank valuations.

The paper makes three contributions. First, it provides among the longest-window evidence (16 years) on ESG disclosure and bank performance in a frontier market, documenting a pronounced upward trend in disclosure alongside persistent heterogeneity across banks. Second, by jointly modeling accounting and market performance, it uncovers a disclosure–performance asymmetry that reconciles the mixed findings of prior work and speaks directly to the maturity of emerging capital markets. Third, it offers actionable implications for banks, regulators, and investors seeking to align transparency with value creation. The remainder of the paper proceeds as follows. Section 2 sketches the institutional background; Section 3 develops the theory and hypotheses; Section 4 describes data and methods; Section 5 reports results; Section 6 discusses them; and Section 7 concludes.

2. Institutional Background: ESG and the Vietnamese Banking Sector

Vietnam’s banking system is the backbone of a bank-based financial system in which firms and households remain heavily reliant on bank credit. Over the past decade, sustainability has become a policy priority for the sector. In 2015, the State Bank of Vietnam issued Directive No. 03/CT-NHNN to promote green credit and the management of environmental and social risks in lending. The Ministry of Finance’s Circular No. 96/2020/TT-BTC required listed entities to publish sustainability-related information—covering greenhouse-gas emissions, resource use, and social impact—within their disclosure obligations. These measures were reinforced by Vietnam’s net-zero-by-2050 pledge at COP26 and by the national green-growth strategy for 2021–2030.

Despite this momentum, ESG disclosure by Vietnamese banks remains largely voluntary in substance and uneven in practice. Independent assessments by the Fair Finance initiative and subsequent academic work document steady but heterogeneous progress, with governance items disclosed more fully than environmental and social ones, and with large, listed banks far ahead of smaller peers (Loan et al., 2024; Thich & Hang, 2023). Only a subset of leading banks has begun aligning reports with international frameworks such as GRI and the ISSB’s IFRS S1/S2 standards. This combination of rising policy pressure, voluntary disclosure, and wide cross-bank dispersion makes Vietnam an informative setting in which to test whether disclosure is financially consequential.

3. Theoretical Framework and Hypothesis Development

3.1. Theoretical Underpinnings

The expected link between ESG disclosure and financial performance rests on several complementary theories. Stakeholder theory (Freeman, 1984; Friedman & Miles, 2002) holds that firms create long-run value by meeting the expectations of a broad set of stakeholders; for banks, whose franchise depends on the trust of depositors, borrowers, and regulators, transparent ESG reporting signals responsiveness to these constituencies and helps secure their continued support, and voluntary disclosure is itself an application of stakeholder management to corporate reporting (Roberts, 1992). Signaling theory (Spence, 1973; Akerlof, 1970) and the disclosure literature (Verrecchia, 2001; Cormier & Magnan, 2015) emphasize that, under information asymmetry, credible voluntary disclosure conveys private information about quality and prospects, thereby lowering the uncertainty and the returns demanded by outside claimants (Myers & Majluf, 1984). Legitimacy theory (Suchman, 1995; Deegan, 2014; Ntim & Soobaroyen, 2013) frames disclosure as a means of demonstrating conformity with societal norms and preserving the firm’s “license to operate,” mitigating reputational and regulatory risk; relatedly, finance itself has been argued to act as a driver of corporate social responsibility (Scholtens, 2006). Finally, agency theory (Jensen & Meckling, 1976) offers a competing view: managers may over-invest in visible ESG activities and reporting to advance private reputational agendas, so that disclosure could be value-neutral or even value-destroying. The net effect is therefore an empirical question.

3.2. ESG Disclosure and Accounting-Based Performance

For accounting performance, the balance of theory and evidence points to a positive association. By reducing information asymmetry, ESG disclosure can lower funding and monitoring costs, broaden and stabilize the deposit and investor base, strengthen customer loyalty, and improve the management of environmental and social risks embedded in the loan book—channels that feed directly into operating profitability (Wu & Shen, 2013; Cornett et al., 2016; Esteban-Sanchez et al., 2017). Empirically, positive links between sustainability engagement or disclosure and bank profitability have been documented across developed and developing markets (Buallay, 2019; Buallay et al., 2021; Menicucci & Paolucci, 2023; Gangwani, 2024), and Shakil et al. (2019) find that environmental and social performance improves the financial health of emerging-market banks; comparable positive evidence is reported for South Africa and Morocco (Khlif et al., 2015; Chininga et al., 2024) and for Borsa Istanbul (Saygili et al., 2022); disclosure has likewise been tied to competitive advantage and performance in Malaysia (Mohammad & Wasiuzzaman, 2021), although evidence within the banking sector is not uniform (Soana, 2011). ESG engagement has likewise been linked to lower bank risk and greater stability (Di Tommaso & Thornton, 2020; Chiaramonte et al., 2022), reinforcing the channel from disclosure to sounder, more profitable operations. In the Vietnamese context, Loan et al. (2024) report that ESG—and especially its environmental and governance pillars—raises bank profitability, with the strongest effect on ROE, while related studies find positive CSR–performance links among Vietnamese banks (Bui, 2021; Tran et al., 2021; Thich & Hang, 2023) and lower risk for socially responsible banks (Nguyen & Nguyen, 2020). Although some studies report negative or insignificant effects (Forgione et al., 2020; Bătae et al., 2020)—typically attributed to the short-run costs of ESG implementation—the weight of evidence, and the mechanisms above, support a positive expectation for internal, accounting-based performance. We therefore hypothesize:

H1. Voluntary ESG disclosure is positively associated with the accounting-based financial performance (ROA and ROE) of commercial banks.

3.3. ESG Disclosure and Market-Based Performance

Whether disclosure is impounded into market value is less clear-cut, particularly in emerging markets. In efficient, ESG-attentive markets, transparency should attract sustainability-oriented and institutional investors, deepen liquidity, and raise valuation multiples such as Tobin’s Q and price-to-book (Eccles et al., 2014; Aydoğmuş et al., 2022; Wong et al., 2021). However, emerging capital markets are frequently characterized by retail-dominated ownership, short investment horizons, and limited ESG rating infrastructure, so that non-financial disclosure may not be fully or promptly reflected in prices (Naeem et al., 2022); the valuation effect is further conditioned by the institutional and cultural context in which investors operate (Matthiesen & Salzmann, 2017). Evidence is correspondingly mixed: some studies find ESG raises firm value in Asian and emerging settings, while others find weak or insignificant market effects. Given the developing state of Vietnam’s stock market and the still-nascent demand for ESG information among its investors, the market response is theoretically ambiguous. We state a directional but tentative hypothesis:

H2. Voluntary ESG disclosure is positively associated with the market-based financial performance (Tobin’s Q and price-to-book) of commercial banks.

Figure 1 summarizes the conceptual framework: voluntary ESG disclosure is expected to affect accounting and market performance through the information and trust mechanisms of stakeholder, signaling, and legitimacy theory, conditional on a set of bank-level controls.

4. Data and Methodology

4.1. Sample and Data

The sample comprises Vietnamese commercial banks observed over 2010–2025, yielding an unbalanced panel. ESG disclosure data are hand-collected from banks’ annual reports and standalone sustainability reports through content analysis; financial and ownership data are drawn from audited financial statements and market databases. Market-based measures (Tobin’s Q and price-to-book) are available only for listed banks, resulting in a smaller sub-sample for those models. Continuous variables are winsorized to mitigate the influence of outliers. The estimation samples contain 491 bank-year observations (38 banks) for the accounting models and 289 observations (27 banks) for the market models.

4.2. Variable Measurement

The dependent variables are two accounting measures—return on assets (ROA) and return on equity (ROE)—and two market measures—Tobin’s Q and the price-to-book ratio (P/B). The variable of interest, ESG, is a voluntary disclosure index scored from 0 to 100 through content analysis across the environmental, social, and governance pillars, with higher values denoting more comprehensive disclosure. Following the bank-performance and ESG literature (Wu & Shen, 2013; Buallay et al., 2021; Loan et al., 2024; Petria et al., 2015; Ongore & Kusa, 2013), we control for bank age (TIME), leverage (LEV), liquidity (CR), size (LNSIZE), board independence (BOARDI), board gender diversity (FEMALE), state ownership (STATE), foreign ownership (FORP), CEO short tenure (NEW), CEO qualification (MBAD), media pressure (MEDIA), standalone sustainability report (FORM), and Big-4 audit (AUDIT), the latter capturing the role of audit quality in the ESG–performance relationship (Zahid et al., 2022). Table 1 defines all variables.

4.3. Empirical Model

To test H1 and H2, we estimate the following panel model separately for each performance measure:

PERF^it = β⁰ + β¹ ESG^it + Σ β^k CONTROLS^it + μⁱ + ε^it

where PERF is one of ROA, ROE, Tobin’s Q, or P/B for bank i in year t; ESG is the voluntary disclosure index; CONTROLS is the vector of bank-level controls in Table 1; μ captures unobserved bank heterogeneity; and ε is the idiosyncratic error.

4.4. Estimation Strategy

For each dependent variable we select among pooled OLS, fixed-effects (FEM), and random-effects (REM) estimators using the F-test (pooled vs. FEM), the Breusch–Pagan LM test (pooled vs. REM), and the Hausman test (FEM vs. REM). We then test for groupwise heteroskedasticity (modified Wald) and first-order autocorrelation (Wooldridge), and we correct for detected defects by estimating the selected models with bank-clustered robust standard errors, which yield valid inference in the presence of both heteroskedasticity and within-bank serial correlation. Because the disclosure index is highly correlated with bank size and age, we assess multicollinearity using variance inflation factors and re-estimate a reduced-form specification (excluding size and age) as a robustness check. Finally, because the ESG–performance relationship may be endogenous—through reverse causality (more profitable banks can afford more disclosure), dynamic persistence of performance, and unobserved heterogeneity—we estimate a dynamic two-step system GMM model (Arellano & Bond, 1991; Arellano & Bover, 1995; Blundell & Bond, 1998; Roodman, 2009). The lagged dependent variable and ESG are treated as endogenous and instrumented with their own lags; the instrument set is collapsed to limit proliferation; Windmeijer-corrected standard errors are used; and instrument validity is assessed via the Hansen J-test and the Arellano–Bond AR(2) test.

5. Results

5.1. Descriptive Statistics

Table 2 reports descriptive statistics. The mean voluntary ESG disclosure score is 40.76 (out of 100), with a standard deviation of 26.39 and a range spanning the full 0–100 interval, indicating that disclosure is, on average, moderate and highly dispersed across banks. The trend over time is strongly upward—rising from roughly 35% in 2010 to about 66% in 2025 (Figure 2)—yet the persistent dispersion signals that transparency remains uneven. Accounting performance is modest (mean ROA 0.75%, ROE 14.24%), consistent with the thin margins and high leverage of banking (mean leverage 0.91). Market measures average near unity for Tobin’s Q (1.03) and 1.40 for P/B.

5.2. Correlation and Multicollinearity

Pairwise correlations (untabulated for brevity) are mostly below 0.8 in absolute value. The notable exceptions involve the disclosure index: ESG is strongly correlated with bank age (0.81) and size (0.72), reflecting that older and larger banks disclose more. Variance inflation factors confirm this concern—ESG (24.9) and size (10.8) exceed the conventional threshold of 10, while the mean VIF is 4.8. We therefore interpret the ESG coefficient cautiously and provide a reduced-form robustness test in Section 5.5.

5.3. Model Selection and Diagnostics

For ROA and ROE, the Breusch–Pagan LM test does not reject a zero variance of the random component (p = 1.000) whereas the F-test rejects pooling (p < 0.05), so the fixed-effects estimator is selected. For Tobin’s Q and P/B, both tests are significant and the Hausman test does not reject the random-effects specification (p = 0.166 and 0.079, respectively), so the random-effects estimator is selected. The modified Wald test indicates groupwise heteroskedasticity in all four models (p < 0.001); the Wooldridge test detects first-order autocorrelation for the market models but not for ROA and ROE (p = 0.146 and 0.951). All models are therefore estimated with bank-clustered robust standard errors.

5.4. Main Results

Table 3 presents the main estimates. Consistent with H1, voluntary ESG disclosure is positively and significantly associated with ROE (β = 0.0021, p < 0.01): banks that disclose more ESG information earn higher returns on equity. The coefficient on ROA is positive as expected but not significant in the full specification (p = 0.231); as shown below, this reflects multicollinearity rather than the absence of a relationship. In contrast, and consistent with the tentative nature of H2, ESG disclosure is not significantly related to either Tobin’s Q or P/B. This accounting-versus-market asymmetry is visualized in Figure 3. Among the controls, leverage is the most consistent determinant—negative for ROA and ROE but positive for P/B—reflecting the dual role of debt in banking. Bank age and liquidity raise accounting performance; CEO short tenure lowers ROA; media pressure and board gender diversity are positively associated with ROA.

5.5. Robustness: Addressing Multicollinearity

Because ESG is highly collinear with size and age, we re-estimate the ROA and ROE models after excluding LNSIZE and TIME (Table 4). In this reduced-form specification, the ESG coefficient becomes positive and significant at the 1% level for both ROA (β = 0.00017) and ROE (β = 0.00173). Two implications follow. First, the positive effect of voluntary ESG disclosure on accounting performance is robust across both measures once collinearity is mitigated, reinforcing H1. Second, the loss of significance for ROA in the full model is attributable to multicollinearity rather than to the absence of an underlying relationship. The results are thus stable across specifications.

5.6. Endogeneity: Two-Step System GMM

The relationship between ESG disclosure and performance may be endogenous: more profitable banks may have greater resources to invest in disclosure (reverse causality), performance is persistent, and unobserved factors may drive both. To address this, we estimate a dynamic two-step system GMM model for ROA and ROE, treating the lagged dependent variable and ESG as endogenous. Table 5 reports the results. The models are well specified: the number of instruments (12) is well below the number of banks (42); the Hansen test does not reject instrument validity (p = 0.713 and 0.863); and the Arellano–Bond AR(2) test finds no second-order serial correlation (p = 0.200 and 0.299). Crucially, ESG disclosure remains positive and significant for both ROA (β = 0.0003, p < 0.01) and ROE (β = 0.0032, p < 0.05) after controlling for endogeneity. This confirms that the positive disclosure–performance relationship is not merely an artifact of reverse causality and provides strong support for H1.

6. Discussion

Our central finding is a disclosure–performance asymmetry: voluntary ESG disclosure improves banks’ accounting performance—most strongly ROE, and ROA once collinearity is addressed—but is not reflected in their market valuation. The accounting result is consistent with stakeholder and signaling theory: by reducing information asymmetry and demonstrating responsiveness to stakeholders and regulators, disclosure lowers monitoring and funding frictions and strengthens the deposit and investor relationships on which bank profitability rests (Wu & Shen, 2013; Cornett et al., 2016). It aligns closely with Loan et al. (2024), who likewise find the strongest ESG effect on ROE among Vietnamese banks, and with the broader emerging-market evidence of Shakil et al. (2019) and Gangwani (2024). Importantly, the effect survives a dynamic system-GMM estimation that addresses reverse causality, indicating that disclosure plausibly drives performance rather than merely reflecting it.

The market result is more striking and, we argue, economically meaningful. In a fully ESG-attentive market, transparency would be capitalized into Tobin’s Q and P/B; the absence of such an effect suggests that Vietnam’s still-developing, retail-oriented stock market has not yet priced sustainability disclosure. This interpretation is consistent with evidence that market responses to ESG are weaker in emerging and less mature markets (Naeem et al., 2022) and with the limited ESG-rating infrastructure available to Vietnamese investors. The asymmetry helps reconcile the mixed findings in the literature: whether ESG “pays” depends not only on the firm but on which performance lens—operational or market—is applied, and on the maturity of the market doing the pricing.

The control-variable results are equally coherent with banking fundamentals. Leverage depresses book returns yet is rewarded in market valuation, mirroring the dual role of deposits as both a cost and a scale/growth signal (Berger & di Patti, 2006). The positive roles of bank age and liquidity, the drag from CEO instability, and the disciplining effect of media pressure together portray accounting performance as driven by experience, prudent balance-sheet management, and external monitoring.

7. Conclusion and Implications

Using a 16-year panel of Vietnamese commercial banks, this study shows that voluntary ESG disclosure is positively associated with accounting-based performance—robustly so for ROE and, after mitigating multicollinearity, for ROA—but is not yet reflected in market-based valuation. Disclosure is rewarded in the books, not (yet) in the market.

The findings carry concrete implications. For banks, the evidence supplies a business case for treating ESG transparency as a value-relevant investment rather than a compliance burden: they should expand and standardize disclosure, build internal ESG governance and data systems, and align reporting with international frameworks to attract institutional and foreign investors and lower funding costs. For regulators, the persistent dispersion in disclosure and its accounting benefits argue for a standardized, sector-specific ESG-disclosure framework for banks—coupling incentives with a path toward mandatory disclosure of material items—and for investment in ESG-rating infrastructure so that the market can price transparency. For investors, ESG disclosure is a useful signal of managerial quality and operating efficiency; the fact that it is not yet priced implies a potential value opportunity in transparent banks that the market has not fully recognized.

The study has limitations that suggest avenues for future research. First, it focuses on a single industry, precluding cross-industry comparison. Second, the disclosure index captures the presence rather than the verified quality of information and involves coder judgment; weighted indices, third-party ratings, or pillar-level decomposition would refine measurement. Third, collinearity between disclosure, size, and age limits the separation of their effects, and the market sub-sample is small. Finally, although our two-step system GMM estimation mitigates endogeneity from reverse causality and dynamic persistence, identification still relies on internal (lagged) instruments; future work employing external instruments or quasi-natural experiments—and extending the comparison to regional peers—would further strengthen causal inference.